fix(server): fix deepseekv2 loading (#2266)

huggingface · Jul 21, 2024 · f3435ba · f3435ba
1 parent 53ec0b7
commit f3435ba
Showing 1 changed file with 0 additions and 3 deletions.
diff --git a/server/text_generation_server/models/custom_modeling/flash_deepseek_v2_modeling.py b/server/text_generation_server/models/custom_modeling/flash_deepseek_v2_modeling.py
@@ -34,7 +34,6 @@
 from text_generation_server.layers.layernorm import FastRMSNorm
 from text_generation_server.layers.rotary import PositionRotaryEmbedding, get_mscale
 from text_generation_server.utils.import_utils import SYSTEM
-from text_generation_server.utils.log import log_once
 from text_generation_server.utils.weights import Weights
 from torch import nn
 from transformers.activations import ACT2FN
@@ -240,7 +239,6 @@ def __init__(
                     if config.attention_bias
                     else None
                 ),
-                quantize=config.quantize,
             )
             self.q_a_layernorm = FastRMSNorm.load(
                 prefix=f"{prefix}.q_a_layernorm",
@@ -261,7 +259,6 @@ def __init__(
                 if config.attention_bias
                 else None
             ),
-            quantize=config.quantize,
         )
 
         self.kv_a_layernorm = FastRMSNorm.load(